We conducted a prospective observational cohort study to assess the impact of pill taking behavior of HIV infected individuals newly started on the nelfinavir as their first protease inhibitor. Various summary measures of pill taking ("adherence descriptor variables") were defined a priori (e.g., fraction of does missed, duration of maximal dosing interval). Our hy pothesis was that there are differences in a dherence, as defined by these summary statistics, between subjects who did and did not suppress HIV viral loads to below the limit of quantification. Methods: Subjects were recruited from HIV clinics of the University of Pennsylvania Health System and other HIV care sites in Philadelphia. Eligibility criteria included 1) being protease inhibitor naive, 2) being newly started on nelfinavir (either BID or TID), 3) additionally being prescribed at least two nucleoside analog reverse transcriptase inhibitors, at least one of which was new, 4) having a viral load >10,000 copies/ml by PCR or >5000 by bDNA, 5) having any CD4 cell count. Subjects were excluded if they were on dual protease inhibitor regimens. Adherence was assessed using the microelectronic monitoring system (MEMS, APREX) on their nelfinavir bottle only. Subjects were followed for 4 months with the primary outcome being suppression of viral load to below the limit of quantification (400 copies/ml) or not. Results: Of the first 30 subjects enrolled, complete data are available for 20 subjects. A total sample size of 40 subjects is planned. The median age of the subjects was 37 years (range, 27-72), 14 (70%) were male, and 19 (95%) were African-American, 6(30%) were current or former injection drug users, 7(35%) were men who have sex with men. Median initial viral load was 59,000 copies/ml (range 10,000-320,000) median initial CD4 count was 320 cells/cm(3), median initial CD4 percentage was 16% (range, 4%-27%), and 12(60%) were prescribed nelfinavir BID. 14 subjects (70%) achieved the outcome of viral load below detectable at 4 months. Subjects who achieved this outcome had stastically significantly better adherence as defined by several of the adherence descriptor variables. These variables include fraction of doses missed (7% vs. 40%, p=0.004), median duration of dosing interval (13 hours 27 hours, BID group p=0.040; 9 hours vs. 12 hours, TID group p=0.020), fraction of dosing intervals greater than 3 days (0% vs. 1%, p=0.002), and duration of longest dosing interval (2 days vs. 11.5 days, p<0.001). Conclusion: These data demonstrate that individuals who achieve maximal suppression of viral loads have adhered more closely to the prescribed regimen than those who have not. Furthermore, these results demonstrate that several different summary statistics of adherence can be used to demonstrate these differences. Further analyses will be performed to identify which of the adherence descriptor variables are best for classifying an individual as sufficiently adherent or not.